Hazard Ratio
How much faster, over time. A hazard ratio compares the rate an event strikes in two groups — the workhorse of survival analysis, and a sharp lens on churn.
- Term
- Hazard ratio
- Is
- Ratio of event rates over time
- From
- Survival / time-to-event analysis
- HR = 1 means
- No difference between groups
Parts of speech & senses
- A hazard ratio is the ratio of the event rates in two groups in time-to-event analysis, showing how much faster or slower an event like churn occurs in one group. "Annual plans had a churn hazard ratio of 0.6 versus monthly."
What a hazard ratio is
A hazard ratio comes from survival analysis, the branch of statistics that studies not just whether an event happens but when. The 'hazard' is the instantaneous rate at which an event occurs among those still at risk — the chance that a subscriber who has lasted this long cancels in the next moment. A hazard ratio is simply the hazard in one group divided by the hazard in another. A ratio of 1 means the two groups face the event at the same rate. A ratio of 2 means the first group hits the event twice as fast; a ratio of 0.5 means half as fast, a protective effect. The measure was born in medicine — comparing how quickly patients on a treatment versus a placebo experience an outcome — and it is most often estimated with the Cox proportional-hazards model, which produces a single hazard ratio while adjusting for other factors.
For marketers, the natural translation is churn and retention. Think of a customer 'surviving' as long as they stay subscribed, and 'the event' as cancellation. A hazard ratio then answers questions raw churn percentages cannot: do customers who came through a referral churn at half the rate of paid-acquisition customers, adjusting for plan and tenure? A hazard ratio of 0.5 says yes, and quantifies it. The key advantage over a simple churn rate is that survival analysis handles censoring — customers who have not yet churned by the end of your data still contribute what is known about them, rather than being dropped or miscounted. That is vital when many customers are still active. So a hazard ratio lets you compare the speed of churn between cohorts, channels, or plan types over time, using every customer's history, not just the ones who have already left.
Hazard ratio versus churn rate and relative risk
A hazard ratio is easy to confuse with a plain churn rate, but they answer different questions. A churn rate is a snapshot: the share of customers who left over a fixed window, say five percent in a month. It ignores when within that window people left and struggles with customers who joined partway through or have not yet had a chance to churn. A hazard ratio is dynamic: it compares the ongoing rate of churning between two groups across the whole timeline, accounting for how long each customer has been at risk. Where a churn rate might say 'both cohorts ended the quarter at eight percent,' a hazard ratio can reveal that one cohort churned heavily early and the other steadily throughout — a difference in timing that changes what you do about it. The churn rate tells you how many; the hazard ratio tells you how fast, and for whom.
A hazard ratio also differs from its cousins relative risk and the odds ratio, even though all three compare two groups and hover around 1. Relative risk compares the cumulative probability of an event by some fixed point — the share who churned by month six. An odds ratio compares odds rather than probabilities. A hazard ratio compares the rate of the event moment by moment across the entire follow-up, not the total by a single cutoff. The practical consequence is that a hazard ratio uses the timing information the other two discard, and it copes with censored, still-at-risk customers that a simple proportion cannot. It does lean on an assumption — that the ratio of hazards stays roughly constant over time, the proportional-hazards assumption — which is worth checking. When timing and incomplete follow-up matter, as they usually do in retention, the hazard ratio is the sharper instrument.
Using a hazard ratio well
Using a hazard ratio well starts with framing the event and the clock clearly: what counts as the event (cancellation, downgrade, first repeat purchase), when the risk starts, and what data are censored. Fit a survival model — a Cox proportional-hazards model is the common choice — so you can compare groups while adjusting for confounders like plan, tenure, and acquisition channel, rather than comparing raw, mixed populations. Read the number plainly: a hazard ratio above 1 means faster events, below 1 means slower, and always report a confidence interval, because a ratio of 0.8 with an interval spanning 1 is not yet evidence of anything. Check the proportional-hazards assumption; if the ratio clearly changes over time, a single number oversimplifies and you may need a time-varying model. Used this way, the hazard ratio turns retention curves into a clear, adjusted comparison of who is leaving faster.
The failures are mostly misreadings. People treat a hazard ratio as if it were a probability — a ratio of 2 does not mean 'twice as likely to ever churn,' it means the rate of churning is twice as high at any given moment, which is not the same as the eventual total. They ignore censoring by throwing out still-active customers, biasing the comparison toward those who left early. They quote the point estimate and hide the confidence interval, dressing up noise as a finding. And they forget the proportional-hazards assumption, applying one constant ratio to groups whose relative risk actually shifts over the customer lifetime. The discipline is to define the event and the timeline precisely, respect censoring, adjust for confounders, report the interval, and check the assumption — so the hazard ratio measures the difference in event speed it was designed to, and no more.
Synonyms & antonyms
Synonyms
Antonyms
Origin & history
Hazard-based survival analysis and the proportional-hazards model derive from statistician David Cox's 1972 paper 'Regression Models and Life-Tables.'
Etymology: source.
Usage trends
Search interest for this term over the last five years:
Common questions
- What is a hazard ratio?
- The ratio of the event rate in one group to the event rate in another over time, from survival analysis. Above 1 means the event happens faster in the first group, below 1 means slower, and 1 means no difference.
- How is a hazard ratio different from a churn rate?
- A churn rate is a snapshot share who left in a window. A hazard ratio compares the ongoing rate of churning between two groups across the whole timeline, using timing and still-active customers a simple rate ignores.
- How do marketers use hazard ratios?
- To compare how fast different cohorts, channels, or plans churn, adjusting for other factors with a survival model. It reveals differences in the speed and timing of churn that headline churn percentages hide.
Resources & people to follow
- referenceRGM analysis — definitions, senses, and usage verified per term
Curated, non-competitor resources verified per term.
Related training
Disciplines
Areas of marketing where hazard ratio is a core concern: